Ideal Free Distribution Author(S): Mark Moritz, Ian M

Home , Ideal free distribution

Mobile Pastoralists in the Logone Floodplain Distribute Themselves in an Ideal Free Distribution Author(s): Mark Moritz, Ian M. Hamilton, Yu-Jen Chen, and Paul Scholte Source: Current Anthropology, Vol. 55, No. 1 (February 2014), pp. 115-122 Published by: The University of Chicago Press on behalf of Wenner-Gren Foundation for Anthropological Research Stable URL: http://www.jstor.org/stable/10.1086/674717 . Accessed: 25/01/2014 08:42

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp

. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

The University of Chicago Press and Wenner-Gren Foundation for Anthropological Research are collaborating with JSTOR to digitize, preserve and extend access to Current Anthropology.

http://www.jstor.org

This content downloaded from 68.250.191.217 on Sat, 25 Jan 2014 08:42:43 AM All use subject to JSTOR Terms and Conditions Current Anthropology Volume 55, Number 1, February 2014 115

Mobile Pastoralists in the Logone to the same amount of resources. In assessing the distribution Floodplain Distribute Themselves of grazing pressure over available common-pool resources, we use the ideal free distribution as an indicator of manage- in an Ideal Free Distribution ment of these resources in a situation in which pastoralists Mark Moritz, Ian M. Hamilton, Yu-Jen Chen, and are committed to an ethos of open access to common-pool Paul Scholte grazing resources (Moritz et al. 2013). Pastoralists’ use of common-pool grazing resources in the Department of Anthropology, The Ohio State University, Logone floodplain is best described as open access as there 174 West 18th Avenue, Columbus, Ohio 43210-1106, are no institutions that regulate pastoral use of and access to U.S.A. ([email protected])/Department of Evolution, the floodplain (Moritz et al. 2013). On the contrary, pastor- Ecology, and Organismal Biology, Department of Mathe- alists emphatically argue that access is open for everyone, and matics, The Ohio State University, 390 Aronoff Laboratory, every year, after the water recedes, thousands of Arab and 318 West 12th Avenue, Columbus, Ohio 43210, U.S.A./De- FulBe pastoralists from Cameroon, Nigeria, and Niger move partment of City and Regional Planning, The Ohio State with more than 200,000 cattle into the Logone floodplain University, 275 West Woodruff Avenue, Columbus, Ohio (Seignobos and Iye´bi-Mandjek 2000; Zborowski 1996). 43210, U.S.A./c/o Nieuwe Teertuinen 12C, Amsterdam, LV Open access to common-pool resources has been equated with 1013, the Netherlands. This paper was submitted 7 IX 12, the tragedy of the commons (Hardin 1968). However, critics accepted 22 VII 13, and electronically published 18 XII 13. have argued that Hardin confused commons with open access CAϩ Online-Only Material: Supplement A or unmanaged common-pool resources (Ciriacy-Wantrup and Bishop 1975; McCay and Acheson 1987; Ostrom 1990). More- We examined whether mobile pastoralists in the Logone over, Ostrom (1990) has shown that common-pool resources floodplain of Cameroon distribute themselves according to can be managed sustainably as common property regimes the ideal free distribution (IFD), which predicts that the num- using game-theoretical models and a wide range of empirical ber of individuals in each area is proportional to the quality cases of successful commons. The current consensus in the and quantity of resources in each area and that all individuals literature on pastoral systems is that in most cases common- have access to the same amount of resources. We used the pool resources are managed as common property regimes but concept to assess the distribution of grazing pressure over also that open access will lead to a tragedy of the commons available common-pool resources as evidence of a complex (Galvin 2009; Homewood 2008). adaptive system in which the spatial distribution grazing pres- Here we will present evidence in support of our hypothesis sure is adjusted to the distribution of resources through in- that the management system in the Logone floodplain is a dividual decision making and passive coordination of move- self-organizing complex adaptive system (Moritz et al. 2013). ments among individual pastoralists. We used a combination As evidence for management of common-pool resources we of spatial and ethnographic approaches to study the distri- expect to see that the distribution of grazing pressure is pro- bution of resources and mobile pastoralists in the Logone portional to the distribution of resources within the floodplain floodplain in 5 successive years and found evidence for an as in an ideal free distribution. IFD in 3 years (2008–2009 and 2012) and an approximation of an IFD in years in which pastoralists were terrorized by The Ideal Free Distribution Model armed bandits (2010) and the government reestablished security (2011). The findings support our hypothesis that there The ideal free distribution is a simple and elegant game the- is a self-organizing management system in which pastoralists oretic model that lies at the basis of many other behavioral distribute themselves effectively over the available resources. ecology models that explain the relationship between animal behavior and population dynamics (Fretwell and Lucas 1969; Sutherland 1996). The IFD model predicts how animals In this article we examine pastoral management of common- should distribute themselves over resource habitats or patches. pool grazing resources in a complex social-ecological system. The two main assumptions in the IFD model that give it its Specifically, we examine whether mobile pastoralists in the name are (1) individuals have complete knowledge about the Logone floodplain in the Far North Region of Cameroon resource quality and quantity of each patch (ideal assumption) distribute themselves according to the theory of the ideal free and (2) individuals are free to move to any patch (free as- distribution (IFD; Fretwell and Lucas 1969), which predicts sumption). When those conditions are met, the theory pre- that individuals will settle first in habitats with high-resource dicts that individuals should be distributed such that no in- density and that, at any given time, all individuals have access dividual can gain more resources by moving to a different ᭧ 2014 by The Wenner-Gren Foundation for Anthropological Research. patch. However, empirical studies of fish and birds have not All rights reserved. 0011-3204/2014/5501-0006$10.00. DOI: 10.1086/ always found support for the model (Sutherland 1996). 674717 There are many different versions of the IFD model, but

This content downloaded from 68.250.191.217 on Sat, 25 Jan 2014 08:42:43 AM All use subject to JSTOR Terms and Conditions 116 Current Anthropology Volume 55, Number 1, February 2014 the depletion model is the most appropriate IFD model to Ideal free models do not necessarily alleviate the tragedy examine the pastoral use of forage in the Logone floodplain of the commons. Indeed, in depletion models, depletion may (Hamilton 2010b; Sutherland and Anderson 1993). There are become severe over the whole landscape, as even very poor several predictions that arise from this model. First is that habitats become occupied (Sutherland 1996). However, we occupied habitats have higher resource density than do un- argue that (1) because pastoralists live in an open system in occupied habitats at any point in time, because high-resource which they can move within and out of the floodplain, the habitats are occupied first. Second, the variance in standing number of available habitats becomes very large at large spatial resource quality of occupied habitats should be lower than scales and (2) in the spatiotemporal patterns of resource de- the variance in resource quality of unoccupied habitats, be- pletion and renewal, in which depleted habitats are renewed cause foragers will tend to deplete all occupied habitats to the and the spatial distribution of herders on the floodplain is same quality while, in unoccupied habitats, current resource reset by the onset of the rainy season and flooding, full oc- quality is unaffected by depletion. Finally, there should be a cupancy and severe depletion of all habitats in the limited positive correlation between total resources in a habitat (re- time of the dry season is unlikely. In other words, we expect source density # area) and total number of foragers in a that over a wide range of conditions, individuals can do better habitat. Under a model incorporating only depletion, resource by moving rather than by increasing intensity of exploitation quality per unit area should be equal in all occupied habitats; in a given habitat. We thus expect to find an ideal, free-type however, if habitats vary in area, larger habitats should include distribution when mobile users with complete information, more individuals (Flaxman and deRoos 2006; Fretwell and freedom and ability to move, independent and individual Lucas 1969). decision-making capabilities, and a preference for habitual The concept of the IFD has been primarily used in behavioral campsites have open access to depletable common-pool re- ecology studies to examine habitat selection and foraging strat- sources that are highly variable in space and time. Such a egies of animals (Hamilton 2010a; Kennedy and Gray 1993; distribution will be resistant to strategies that overselect or Sutherland 1996), but it is now also used to study the mobility overstay in certain habitats, because the costs of overusing and settlement patterns of human populations (Kennett, An- these habitats fall most heavily on the individuals that overuse derson, and Winterhalder 2006; Winterhalder et al. 2010), in- them (Hamilton 2010a). cluding mobile pastoralists (Behnke, Davidson et al. 2008; Behnke, Fernandez-Gimenez et al. 2011; Scholte et al. 2006). Study Area and Population Pastoral systems provide an excellent case for examining IFD theory because the mobility and settlement patterns of pas- Two phytogeographic zones characterize the Far North Region toralists are primarily driven by spatiotemporal changes in of Cameroon: Sudanian in the southern grades and Sahelian the distribution of forage and water, even though the human in the Logone floodplain. The Logone floodplain is flooded dimension of pastoral systems means that nonforage con- by the Logone River and its branches from September until straints shape mobility and settlement patterns (Behnke, Fer- November. After the water recedes in December, thousands nandez-Gimenez et al. 2011). of Arab and FulBe pastoralists from Cameroon, Nigeria, and Theoretical and formal models are necessarily much sim- Niger move with more than 200,000 cattle into the floodplain, pler than the complexity of social-ecological systems, and this making it one of the most important dry season grazing lands is why the concept of the IFD is often used as a heuristic in the Lake Chad Basin (Seignobos and Iye´bi-Mandjek 2000). device (Kennett, Anderson, and Winterhalder 2006). How- Many remain there until the start of the rainy season in June, ever, in many ways the social-ecological system of the Logone while others move to other seasonal grazing areas. Pastoralists floodplain is an appropriate setting to use the concept of the find nutritious regrowth and surface water in the floodplain IFD to describe and evaluate the distribution of mobile pas- far into the dry season, when surrounding pastures have dried toralists. Most importantly, the system meets the IFD’s two up. At the start of the rainy season, pastoralists return to main assumptions: first, while pastoralists do not have perfect Diamare´ plains or their respective countries. information, they have reliable information about the suit- The vegetation in the floodplain is relatively homogeneous ability of different habitats through a combination of scouting in terms of forage quantity and quality because of the extreme and social networks; second, access to common-pool grazing flatness of the area resulting in only limited variation in flood- resources in the floodplain is best described as open access ing depth and duration (Scholte 2007). There is only weak in that there are no regulations restricting access, which means coupling between herbivores and vegetation, as the predom- that pastoralists are free to relocate if it is to their advantage inantly perennial vegetation is controlled by flooding depth to do so. We argue that the IFD, and, in particular, the de- and duration and naturally protected against overgrazing be- pletion model, is a useful concept to understand the distri- cause up to two-thirds of the biomass is stored underground. bution of pastoralists in the Logone floodplain. The social- In addition, above-ground biomass is generally inaccessible ecological system, however, does not fit several of the other to livestock during the flooding for 6 months of the year assumptions of ideal free models (see supplement A, available (Scholte 2007). online, for a detailed discussion). Our study area, called Yaayre in Fulfulde, is an approxi-

This content downloaded from 68.250.191.217 on Sat, 25 Jan 2014 08:42:43 AM All use subject to JSTOR Terms and Conditions Moritz et al. Mobile Pastoralists in the Logone Floodplain 117

Figure 1. The study area in the Far North Region of Cameroon. mately 800 km2 section of the Logone floodplain with well- Shinwa). The population consists of approximately 1,000 defined boundaries of the Waza National Park in the west, households divided over approximately 130 camps and in- the Logone River in the east, the irrigated rice fields of SEMRY cludes Suwa Arabs and FulBe, which are subdivided in Jam- in the south, and the village of Zina in the north (see fig. 1). aare, Mare, Alijam, Adanko, and Anagamba groups. These The research site overlaps with the pilot zone of the Waza different FulBe groups are endogamous and have their own Logone Project (1990–2003), which started reflooding of the dialect, cattle breed, houses, and marriage system. pilot zone by opening an old waterway in the embankment Pastoralists set up their camps in koodaande (singular hoo- along the Logone River in 1994 (Loth 2004; Scholte 2005; daande), which refer to sites, that is, the location of herds Scholte et al. 2006). and households, as well as the larger area in which these sites The study population consists of all mobile pastoralists who are located. We refer to the former as campsites and the latter use the Logone floodplain at some point during the year. as camp zones. In the floodplain, camp zones are often located Many of them use the floodplain during the cold dry season in depressions for easy access to water for humans and calves in November and December when the water retreats and then and protection from bush fires. They are also located close either leave for the floodplain further north (Yaayre Woylare) to the three rivers that intersect the floodplain in the east or or the lakeshore pastures of Lake Maga in the south (Ndiyam close to the artificial lakes in the west that were dug by the

This content downloaded from 68.250.191.217 on Sat, 25 Jan 2014 08:42:43 AM All use subject to JSTOR Terms and Conditions 118 Current Anthropology Volume 55, Number 1, February 2014

Table 1. Results from unpaired t-test comparing NDVI values of occupied and unoccupied cells in February, 2008–2012

Occupied cells Unoccupied cells Unpaired t-test Year Avg. Min Max Variance Avg. Min Max Variance t df P- values 2008 .244 .156 .306 .0008 .242 .130 .362 .0017 .662 787 .6620 2009 .261 .165 .362 .0008 .251 .157 .381 .0015 3.924 787 !.0001* 2010 .255 .188 .326 .0008 .242 .182 .378 .0010 5.363 787 !.0001* 2011 .296 .239 .363 .0005 .293 .191 .412 .0013 1.351 787 .1772 2012 .267 .196 .334 .0006 .246 .166 .335 .0008 1.294 787 !.0001* Note. The average normalized difference vegetation index (NDVI ) values in occupied cells are signiﬁcantly higher in 3 of the 5 years. Moreover, the variance is much greater in the unoccupied cells compared to the occupied cells. * Signiﬁcant at the 1% level.

Lake Chad Basin Commission (LCBC). Some camp zones are MODIS MOD13A2 satellite data, which are collected every places where pastoralists stay for an extended period (so- 16 days at 1 km resolution, to calculate the NDVI values for journ), up to 6 months, while in other zones they stay for our study area. While NDVI is not a reliable index for absolute only a few days (transit). Different pastoralists may use the measurements of vegetation, it is very useful as a relative same camp zones for either transit or sojourn. measure of green biomass in the same area at the same time In general, mobile pastoralists in our sample had three (Anyamba and Tucker 2005). sojourn camps each year: in the rainy season (duumol), cold Using ArcGIS software, we divided up our study areas in dry season (daBBunde), and the hot dry season (ceedu). These a grid with 1 km2 cells and calculated the NDVI value for all campsites are referred to, respectively, as ruumirde, daBBirde, 1km2 cells in the study areas for the month of February. We and ceedirde. Pastoralists generally stayed the longest in their used pastoralists’ camp zones as our habitats. To calculate the ceedirde, some up to 6 months. By February pastoralists have NDVI for the camp zones we created buffers with a 1.5 km settled in their ceedirde in the Yaayre, while the ones that have radius around each campsite (using the GPS coordinates from their ceedirde in Ndiyam Shinwa and Yaayre Woylare have our surveys). We combined the buffers that overlapped and left. In the hot dry season, access to water restricts, and pas- shared toponyms into one larger buffer (e.g., all the buffers toralists settle in a limited number of well-established camp from the Cubuna campsites were combined into one Cubuna zones near the rivers that dissect the floodplain or near the buffer). artificial lakes in the western end of the floodplain. They We used a combination of statistical analyses to examine generally stay there until the beginning of the rainy season (in June) or whenever they think that their animals will do whether there was an IFD of mobile pastoralists in the Logone better elsewhere. floodplain (using GraphPad InStat [version 3.0a] and SPSS [version 19] for Macintosh). First, we used descriptive sta- Methods tistics, t-tests, and analysis of variance (ANOVA) to describe and compare the NDVI values for the study area, as well as Data for this article were collected in a longitudinal study of the occupied and unoccupied cells in the different years. Sec- pastoral mobility in the Far North Region of Cameroon ond, we used descriptive statistics to describe the population (2008–2012). We used a combination of spatial and ethno- of mobile pastoralists for the different years (2008–2012). graphic approaches to describe and explain the distribution Third, we used correlations and multiple regressions to ex- of mobile pastoralists in our study area in the Logone flood- amine whether there is a correlation between the NDVI and plain in 5 successive years, 2008–2012. the number of cattle for the camp zones in each year. We also To document the distribution of mobile pastoralists, we calculated for each camp zone the percentage of the NDVI used GPS/GIS technology to map the locations of all pastor- and of cattle (as a percentage of all the camp zones combined alists’ camps in our study area in the floodplain at three different times during the year (November, February, May). for each year) to examine whether there was under use or During these visits, we identified all the households in the overuse of habitats and whether there was a correlation be- camp, the number of herds, and the name of the campsite. tween the properties of each camp zone (e.g., NDVI, number To estimate the distribution of the available resources in of cattle) and under use or overuse. the floodplain we used remote sensing data, in particular the normalized difference vegetation index (NDVI), which is an Results indicator of the quality and quantity of forage. NDVI is one of the most common remote sensed vegetation indices for Here we will describe the results from our study. First, we measuring the plant growth and vegetation cover (Anyamba describe patterns in NDVI and the pastoral population. Sec- and Tucker 2005), including in the assessment of pastoral ond, we will discuss correlations between the NDVI and graz- ecosystems (Butt 2010; Coughenour et al. 2008). We used ing pressure for 5 years (2008–2012).

This content downloaded from 68.250.191.217 on Sat, 25 Jan 2014 08:42:43 AM All use subject to JSTOR Terms and Conditions Moritz et al. Mobile Pastoralists in the Logone Floodplain 119

Table 2. Population of mobile pastoralists in the study area in February, 2008–2012

No. camps No. cattle No. households Households per camp Cattle per household Cattle per camp 2008 36 26,370 296 8.2 89 733 2009 55 36,985 445 8.1 83 672 2010 57 34,270 383 6.7 89 601 2011 54 30,410 318 5.9 96 563 2012 58 21,750 271 4.7 80 374 Average 52 29,957 343 6.8 87 589 Note. The number of cattle per herd are estimated and based on self-reporting, vaccination records, and direct counts as assessed during our earlier studies (Scholte et al. 2006).

NDVI in Camp Zones Changes in Pastoral Population in the Study Area The Logone floodplain is extremely flat, and the vegetation The pastoral population in the Logone floodplain in February is relatively homogeneous, and initially we envisioned the changed from year to year, and this is an indication of how floodplain as one large camp zone. However, analysis of trans- decision making of individual pastoralists results in adjust- humance surveys and GPS data of their camps shows that ment of grazing pressure at the regional level. The pastoral there are distinct camp zones that are occupied year after year, population in the floodplain in February represents about a while other areas are never occupied. third of the mobile pastoralists in our study population (Mo- A visual inspection shows that most of the camp zones are ritz et al. 2013). In February, at the beginning of the hot dry located near depressions and water sources (e.g., rivers and season, another third has moved further north to the flood- the artificial lakes). We also examined whether cells in the plain called Yaayre Woylare, while the remainder has moved camp zones have more grazing resources by comparing the south to the grazing areas surrounding Lake Maga or the NDVI values inside the camp zone buffers (occupied cells) floodplain in neighboring Chad. and outside the buffers (unoccupied cells) for 5 different years The number of camps in our study area in February ranged (2008–2012). We found a number of differences in NDVI from 36 in 2008 to 58 in 2012 with an average of 52 for all between occupied and unoccupied areas, which all support 5 years (see table 2). The number of households also varied our predictions. First, the average NDVI values are higher for considerably from year to year, with 271 in 2012 and 445 in occupied areas in all 5 years, and in 3 years the differences 2009—an average of 343 for all 5 years. The same is true for are statistically significant (see table 1). Second, the distri- the number of cattle, which varied from 21,750 in 2012 and bution of NDVI in the occupied cells passed the normality 36,985 in 2009—an average of 29,957 for all 5 years. test using the Kolmogorov-Smirnov test (P 1.10), but the Moreover, there are no statistically significant correlations unoccupied cells did not (P !.007). Third, the range of NDVI between the number of camps and the households; these values is much larger and the variance is higher in unoccupied numbers vary independently of each other. It is an indication cells than in occupied cells in all 5 years (see table 1). More- of the flexibility of the social system, which is the result of over, if we compare the average NDVI values for the different continuous processes of fission and fusion in the social or- camp zones (total NDVI/ surface area), which is an indicator ganization of camps. Every time pastoralists move, the com- of resource quality per unit area, we found very little variance position of the camp potentially changes. Households from between camp zones (see supplement A for more information one camp can decide to split ways, or households from dif- on intra- and interannual changes in NDVI). ferent camps can decide to set up camp together (or both). Households are autonomous decision-making units, and pas- Table 3. Correlations between NDVI and toralists make independent decisions about where and when number of animals in camp zones, 2008– to move. This indicates that there is considerable flexibility 2012 in the system and that there are no major social constraints in the movements of pastoralists (see supplement A for more Linear correlation information on change and continuity in pastoral popula- Year rN P tions). 2008 .8270 16 !.0001* 2009 .7431 17 .0006* Matching NDVI and Animal Numbers 2010 .4394 17 .0776 We further examined whether there was evidence for an ideal 2011 .4648 18 .0520 2012 .8558 16 !.0001* free distribution in the Logone floodplain by testing whether there was a positive correlation between the total NDVI and Note. In all 5 years there are strong correlations between the total normalized difference vegetation index the total number of animals per camp zone for the 5 different (NDVI) and the number of animals in camp zones. years. We found strong correlations between the NDVI and * Significant at the 1% level. number of animals per camp zone buffer in 3 of the years

This content downloaded from 68.250.191.217 on Sat, 25 Jan 2014 08:42:43 AM All use subject to JSTOR Terms and Conditions 120 Current Anthropology Volume 55, Number 1, February 2014

Table 4. Correlations between overuse and under use and Interference Due to Insecurity number of animals in camp zones, 2008–2012 While we found strong support for our predictions of the Linear correlation Matching in zones distribution of mobile pastoralists in the Logone floodplain in 2008–2009 and 2012, the evidence is less strong in 2010– Year rN POver Under 2011. One explanation for the absence of strong evidence for ! 2008 .8423 16 .0001* 6 10 an IFD in these years may be the insecurity in the Far North 2009 .6705 17 .0032* 5 12 2010 .8758 17 !.0001* 6 11 Region during those years. Mobile pastoralists were terrorized 2011 .7461 18 .0004* 6 12 by armed bandits (called coupeurs de route), who kidnapped 2012 .9174 16 !.0001* 6 10 children for ransom (Moritz and Scholte 2011). The kidnap- Note. There are strong correlations between the degree of overuse and pings started in the rainy season of 2009 and continued for under use and the number of animals in camp zones. Camp zones with more than a year. In October 2009, when pastoralists moved higher numbers of animals are more likely to be overused. from their rainy season camp zones to the Logone floodplain, * Significant at the 1% level. six children were taken hostage, and 31 million francs (approximately $61,600) was paid in ransoms (OPEN 2009). The (2008, 2009, 2012) and near significant correlations in the bandits were well organized and heavily armed, and they did other 2 years (2010, 2011; see table 3). The fact that the not hesitate to kill. They operated in small groups of three correlation is positive in all 5 years suggests that there is a to five, and are thought to be mostly former rebels, army consistent pattern between NDVI and the number of animals deserters, and outcasts from pastoral groups in the Chad Ba- across camp zones. Moreover, if we remove the outlier camp sin. Livestock traders and pastoralists were their main targets, zone Lugge with 5,560 animals in 2010, there is a statistically and they operated in the bush and on secondary roads linking significant correlation (r p 0.6704, n p 16, P p .0045). the weekly rural markets (Moritz and Scholte 2011). As a We also examined whether there were any patterns in over- result, many pastoralists changed their annual transhumance use and under use, that is, whether the percentage of NDVI orbit and stayed away from their habitual camp zones if they per camp zone buffer (as a percentage of the NDVI of all the considered the risks of kidnappings too high. Instead, they camp zone buffers) was equal to the percentage of animals opted for staying closer to others to minimize the risks of per camp zone buffer (as a percentage of the animals in all attacks. Many pastoralists in those years talked about leaving the camp zone buffers) in each of the years. If the percentage Cameroon altogether and leaving for Nigeria or Chad (and of animals was higher than the percentage of NDVI, it is a few did). considered a case of overuse (more animals than resources). While we have collected qualitative and quantitative data If the percentage of animals was lower than the percentage on insecurity in the Far North Region (Moritz and Scholte of NDVI, it is considered under use (fewer animals than 2011), there is no clear evidence for the impact of insecurity resources). We found that the number of camp zones that in our study area in terms of the number of animals and the were overused stayed more or less the same across the years distribution of the camps, which do not seem to be different (see table 4). Although it was not always the same camp zones from other years. However, the composition of the pastoral that were overused (see supplement A for maps of the dis- population changed; pastoralists that previously stayed further tribution of pastoralists in relation to NDVI), the camp zones north in Yaayre Woylare during the hot dry season now stayed that were used every year (or more frequently) were more in our study area. What is different is the weak correlation likely to be overused than the zones that were not used every between NDVI and the number of animals during the years year (Spearman r p 0.7030, n p 32, P !.0001). In addition, of insecurity in the Far North Region, suggesting that inse- the camp zones with more animals were more likely to be curity had an impact on how pastoralists distributed them- overused than the camps with fewer animals (see table 4). selves over the available grazing resources. This is not surprising, because the difference between NDVI values in camp zone buffers vary relatively little (see table 1), Discussion while the number of animals per camp zone varies consid- We derived three predictions from the depletion and inter- erably (see table 5). ference models of the IFD. First, occupied areas should have higher resource density than do unoccupied areas at any point Table 5. Number of animals per camp zone, 2008–2012 in time. Second, the variance in resource quality of occupied areas should be lower than the variance in resource quality Year Average SD Min Max Total for study area of unoccupied areas. Third, there should be a positive cor- 2008 1,648 1,066 255 3,810 26,370 relation between total resources and total animal numbers in 2009 2,175 1,458 390 5,735 36,985 a camp zone buffer. Our findings lend support for all three 2010 2,016 1,354 425 5,560 34,270 predictions. First, the resource density in occupied areas is 2011 1,689 1,142 255 4,145 30,410 significantly higher in 3 of the 5 years. Moreover, we find that 2012 1,357 1,180 210 3,950 21,715 areas that are occupied year after year have significantly more

This content downloaded from 68.250.191.217 on Sat, 25 Jan 2014 08:42:43 AM All use subject to JSTOR Terms and Conditions Moritz et al. Mobile Pastoralists in the Logone Floodplain 121 resources. Second, in all 5 years we find that there is lower natural factors that influence livestock distribution and rangeland deser- tification in Turkmenistan. In NATO Science for Peace and Security Series. variance in resources in occupied areas than in unoccupied R. Behnke, ed. Pp. 141–170. Dordrecht: Springer. areas. Third, we find a positive correlation between the total Behnke, R. H., M. E. Fernandez-Gimenez, M. D. Turner, and F. Stammler. resources and total numbers of cattle in a camp zone buffer 2011. Pastoral migration: mobile systems of livestock husbandry. In Animal in all 5 years. The results do not allow us to make inferences migration: a synthesis. E. J. Milner-Gulland, J. M. Fryxell, and A. R. E. Sinclair, eds. Pp. 144–171. Oxford: Oxford University Press. about which exact combination of formal depletion and in- Butt, B. 2010. Pastoral resource access and utilization: quantifying the spatial terference models best describes the distribution of mobile and temporal relationships between livestock mobility, density and biomass pastoralists in the Logone floodplain, but there is clearly sup- availability in southern Kenya. Land Degradation and Development 21(6): 520–539. port for an ideal, free-type distribution. Ciriacy-Wantrup, S. V., and R. C. Bishop. 1975. “Common property” as a Using behavioral ecology models like the IFD to study the concept in natural resources policy. Natural Resources Journal 15:713–727. dynamics of complex social-ecological systems comes with its Coughenour, M., R. Behnke, J. Lomas, and K. Price. 2008. Forage distributions, own set of challenges. First, social-ecological systems likely range condition, and the importance of pastoral movement in Central Asia—a remote sensing study. In NATO Science for Peace and Security Series. do not meet all the assumptions underlying the models (see R. Behnke, ed. Pp. 45–80. Dordrecht: Springer. supplement A for a detailed discussion). Second, human be- Flaxman, S. M., and C. A. deRoos. 2006. Different modes of resource variation havior is more complex than habitat selection or forage strat- provide a critical test of ideal free distribution models. Behavioral Ecology and Sociobiology 61:877–886. egies in the models, which is, of course, true for all theoretical Fretwell, S. D., and J. H. J. Lucas. 1969. On territorial behavior and other models. However, the advantage of using models, like the factors influencing habitat distribution in birds. Acta Biotheoretica 19:16– IFD, to study complex social-ecological systems lies in its 36. Galvin, K. A. 2009. Transitions: pastoralists living with change. Annual Review simplicity. In this case, for example, it helps us to understand of Anthropology 38:185–198. the spatiotemporal complexity of pastoral mobility. Pastoral Hamilton, I. M. 2010a. Foraging theory. In Evolutionary behavioral ecology. mobility in arid and semiarid lands can be explained at the D. Westneat and C. Fox, eds. Pp. 177–193. Oxford: Oxford University Press. macro level in terms of spatiotemporal patterns in availability ———. 2010b. Habitat selection. In Encyclopedia of animal behavior.M.D. Breed and J. Moore, eds. Pp. 38–43. Oxford: Academic Press. in forage and water, which in turn is driven by seasonal pat- Hardin, G. 1968. The tragedy of the commons. Science 162:1243–1248. terns in rainfall (see also McCabe 2004). The question that Homewood, K. 2008. Ecology of African pastoralist societies. Oxford: James remains, however, is how exactly the IFD emerges. We think Currey. Kennedy, M., and R. D. Gray. 1993. Can ecological theory predict the dis- that it emerges from a dynamic process in which open access, tribution of foraging animals? A critical analysis of experiments on the ideal habitual movements, participation in an information-sharing free distribution. Oikos 68:158–166. network, and pastoralists’ independent decision making are Kennett, D. J., A. Anderson, and B. Winterhalder. 2006. The ideal free dis- all critical factors (Moritz et al. 2013) and plan to examine tribution, food production, and the colonization of Oceania. In Behavioral ecology and the transition to agriculture. D. J. Kennett and B. Winterhalder, that hypothesis using an agent-based model. eds. Pp. 265–288. Berkeley: University of California Press. Loth, P., ed. 2004. The return of the water. Gland, Switzerland: IUCN. McCabe, J. T. 2004. Cattle bring us to our enemies: Turkana ecology, politics, and raiding in a disequilibrium system. Ann Arbor: University of Michigan Acknowledgments Press. McCay, B. J., and J. M. Acheson. 1987. The question of the commons: the culture We would like to thank the mobile pastoralists in the Far and ecology of communal resources. Arizona Studies in Human Ecology. North Region of Cameroon for participating in this study. Tucson: University of Arizona Press. Moritz, M., and P. Scholte. 2011. Ethical predicaments: advocating security We also want to thank Centre d’Appui a la Recherche et au for mobile pastoralists in weak states. Anthropology Today 27:12–17. Pastoralisme (CARPA) for research support in the field. This Moritz, M., P. Scholte, I. M. Hamilton, and S. Kari. 2013. Open access, open research has been supported by the National Science Foun- systems: pastoral management of common-pool resources in the Chad Ba- dation (BCS-0748594), the National Geographic Society sin. Human Ecology 41(3):351–365. OPEN. 2009. Document de pre´sentation des faits majeurs relatifs a`l’insećurite´ (8306-07), and the College of Social and Behavioral Sciences dont ont´te e ´ objet les communaute´s des pasteurs dans les plaines se`ches et and the Department of Anthropology at Ohio State University humides de la Re´gion de l’Extreˆme Nord du Cameroun. Maroua, Cameroon: (OSU). We would like to thank the Ministry of Scientific l’Observatoire du Pastoralisme dans la Province de l’Extreˆme Nord (OPEN). Ostrom, E. 1990. Governing the commons: the evolution of institutions for col- Research and Innovation (MINRESI), the Wildlife College in lective action. Cambridge: Cambridge University Press. Garoua, and the Higher Institute of the Sahel at the University Scholte, P. 2005. Floodplain rehabilitation and the future of conservation and of Maroua for granting research permission and research af- development: adaptive management of success in Waza-Logone, Cameroon. PhD dissertation, Leiden University. filiation (2008–2012). Yu-Jen Chen made the figures and ———. 2007. Maximum flood depth characterises above-ground biomass in maps. We would like to thank the reviewers for their critical African shallowly flooded grasslands. Journal of Tropical Ecology 23:63–72. comments and helpful suggestions. Scholte, P., S. Kari, M. Moritz, and H. Prins. 2006. Pastoralist responses to floodplain rehabilitation in Northern Cameroon. Human Ecology 34:27–51. Seignobos, C., and O. Iye´bi-Mandjek, eds. 2000. Atlas de la Province Extreˆme- References Cited Nord Cameroun. Paris: IRD & MINREST. Sutherland, W. J. 1996. From individual behaviour to population ecology.Ox- Anyamba, A., and C. J. Tucker. 2005. Analysis of Sahelian vegetation dynamics ford: Oxford University Press. using NOAA-AVHRR NDVI data from 1981–2003. Journal of Arid Envi- Sutherland, W. J., and C. W. Anderson. 1993. Predicting the distribution of ronments 63:596–614. individuals and the consequences of habitat loss: the role of prey depletion. Behnke, R., G. Davidson, A. Jabbar, and M. Coughenour. 2008. Human and Journal of Theoretical Biology 160:223–230.

This content downloaded from 68.250.191.217 on Sat, 25 Jan 2014 08:42:43 AM All use subject to JSTOR Terms and Conditions 122 Current Anthropology Volume 55, Number 1, February 2014

Winterhalder, B., D. J. Kennett, M. N. Grote, and J. Bartruff. 2010. Ideal free settlement of California’s Northern Channel Islands. Journal of Anthropo- logical Archaeology 29:469–490. Zborowski, I. D., ed. 1996. Atlas d’e´levage du bassin du Lac Tchad. Montpellier, France: CIRAD-CTA.

This content downloaded from 68.250.191.217 on Sat, 25 Jan 2014 08:42:43 AM All use subject to JSTOR Terms and Conditions ᭧ 2014 by The Wenner-Gren Foundation for Anthropological Research. All rights reserved. DOI: 10.1086/674717

Supplement A from Moritz et al., “Mobile Pastoralists in the Logone Floodplain Distribute Themselves in an Ideal Free Distribution” (Current Anthropology, vol. 55, no. 1, p. 115)

Applying the IFD Model in Complex Social-Ecological Systems We argue that the ideal free distribution (IFD) model, and, in particular, the depletion model, is a useful concept to understand the distribution of pastoralists in the Logone floodplain. The social-ecological system does, however, not fit several of the other assumptions of ideal free models. This poses a number of challenges in the operationalization of the theoretical model so that the IFD can be used to study the distribution of mobile pastoralists in the Logone floodplain. First, cattle are organized in herds, so that movement decisions of individual cattle are not independent of those of other cattle. However, since we are studying the social-ecological system at a relatively large scale—on average, 343 herds and 17 habitats with large amounts of grazing resources—it is not problematic that decisions about movements are made at the level of the herd (rather than individual animals). Second, pastoralists have no perfect knowledge about the suitability of all the patches. We argue that pastoralists have complete information, that is, sufficient knowledge to predict with a certain probability the patch conditions, instead of perfect information, that is, knowing the exact patch conditions (Kelly 1995). Cressman and Krivan (2006) and Cantrell, Cosner, and Lou (2008) modeled the evolutionary stability of movement rules that lead to ideal free distributions. These models predict that rules that achieve the IFD can yield higher payoffs than alternative rules, even when individuals only have information about their current patch and one other patch. Thus, even under limited information, ideal free habitat use may result. Third, there are no distinct identifiable patches or habitats. Instead, habitats are defined by distribution of pastoralists in camp zones, which may overlap (see also Arditi and Dacorogna 1988). Cantrell, Cosner, and Lou (2008) modeled habitat selection along a gradient of habitat suitability, rather than discrete patches. They found that movement rules using local information could approximate the IFD. Fourth, herds are not entering the floodplain and making a habitat selection one at a time. Instead, all mobile pastoralists enter the floodplain en masse when the floods recede. This makes it difficult to assess the suitability of the different habitats (because the cattle densities of the habitats are unknown), were it not that pastoralists usually go to their habitual camp zones, which allows pastoralists to make predictions about the expected distribution of other pastoralists and thus the suitability of the different habitats. Fifth, even though forage and water are the most important factors in settlement decisions, other factors also affect settlement decisions. In other words, habitat quality is also shaped by nonforage constraints like access to markets and insecurity (Behnke et al. 2011; Winterhalder and Kennett 2006). This makes it difficult to assess suitability of habitats in complex social-ecological systems. Access to water and foraging resources is obviously important, but it is less clear whether and how nonforage factors should be weighed. Sixth, one of the assumptions in behavioral ecology is that individuals make strategic decisions that optimize their fitness, and model outcomes are measured in terms of individual fitness. However, measuring fitness, or proxies like caloric returns, is difficult in this pastoral system. We tried using the cattle condition scoring technique (Nicholson and Butterworth 1986), but that turned out to be unfeasible. Instead, we use the concept of the IFD to measure the impact of mobile pastoralists on the environment with the assumption that if grazing pressure is proportionally distributed over the available resources, it is evidence of management of open access. Seventh, travel times between habitats are non-negligible. It may take herds 2 or more days to get from one end of the floodplain to another. Moreover, during these movements animals expend energy and have less time to graze (which they make partially make up by grazing longer at night during movement days). Eighth, pastoralists do not have similar preferences and assessment of habitats; their annual transhumance shapes and is shaped by habitual preferences for habitats on these orbits. The IFD model assumes that all individuals have the same preferences and assessment of habitats. That is clearly not the case in the Logone floodplain, where pastoralists show preference for specific camp zones. Ninth, the distribution of pastoralists in the floodplain may be affected by Allee’s principle, which explains that a patch’s suitability initially is low when population densities are low and increases due to increasing population densities

1 Supplement A from Moritz et al., Mobile Pastoralists in the Logone Floodplain

(Allee et al. 1949; Sutherland 1996). Allee’s principle may play a role in the floodplain, as the perceived suitability of a camp zone increases initially when there are more (not fewer) pastoralists because pastoralists who camp alone or with only a few households are at higher risk of cattle theft. The challenges notwithstanding, we think that the IFD is an appropriate model to describe and explain habitat selection of mobile pastoralists in the hot dry season in the Logone floodplain. We also argue that the IFD is an indicator of a self- organizing system of management of common-pool grazing resources in the Logone floodplain.

Results Normalized Difference Vegetation Index in the Study Area Pastoral mobility is driven by seasonal variations in rainfall, which results in spatiotemporal variation in pastoral resources: forage and water. FulBe pastoralists in the Far North region talk about four distinct seasons when they discuss their mobility patterns. The rainy season (duumol) runs from July through September with the biggest rains in July and August. It is followed by the cold dry season, (daBBunde) from October to January and the hot dry season (ceedu) from February to May. June is a transition season (seeto) from dry to rainy season. The change of seasons varies from year to year; that is, the hot dry season does not always start in February—sometimes it starts in March. The change in normalized difference vegetation index (NDVI) values in the floodplain from a high of 0.7 to a low of 0.2 is a good reflection of the seasons described above (see fig. A1). NDVI values reach the highest point in the beginning of August (0.7–0.8 after the period of the big rains called loddo), then values start to decline, only to increase slightly again in September due to the flooding. Then at the beginning of October (at the beginning of the cold dry season) the values decline, and they reach the lowest value in April (0.2 in the middle of the hot dry season) and again start to increase at the end of May or beginning of June (depending on the start of the rains).

Figure A1. Average NDVI values in the study area, 2008–2012. The rainy season runs from July through September with the biggest rains in July and August. It is followed by the cold dry season from October to January and the hot dry season from March to May. June is a transition season from dry to rainy season.

The changes in NDVI capture well the general seasonal and flooding dynamics described for the Logone floodplain (Delclaux et al. 2010; Jung et al. 2011; Scholte 2005) that drive the movements of mobile pastoralists. Pastoralists leave the rainy season grazing areas (ruumirde) in the beginning of October and enter the floodplain by November, where they settle in their dry season campsites (ceedirde) by January or February and leave by May or June following the rain clouds.

Change and Continuity in Pastoral Populations There is some continuation in occupation of the different camp zones over the years, but the changes from year to year in who occupies what camp zone are considerable (see tables A1 and A2). In the 5 years we documented the distribution of

2 Supplement A from Moritz et al., Mobile Pastoralists in the Logone Floodplain mobile pastoralists in the study area in February, we recorded 131 unique camp leaders. Only 35 of these camps (or camp leaders) had their ceedirde in the study area 3 or more years, while 96 were in the study area in February for only 1 or 2 years (see table A1). Moreover, in the 5 years pastoralists stayed in 35 unique camp zones in the study area in February. Only 16 of these camp zones were occupied for 3 years or more, while 19 camp zones were occupied for only 1 or 2 years in February (see table A2). Thus, the pastoral population and the camp zones that they occupied changed considerably from year to year. However, our results show that this had no major effect on matching of resources with animal numbers.

Table A1. Continuation in occupation—camp leaders, 2008–2012 Same zone Years in the study area Number of camps Same zone in all years in most of years 51038 412311 313710 2261010 170NA NA Total 131 21 39 Note. NA p not applicable.

Table A2. Continuation in occupation—camp zones, 2008–2012 Total number of camps Average number of Years occupied Number of camp zones over 5 years camps per year 5 7 121 3.5 4 5 58 2.9 3 4 34 2.8 2 4 14 1.8 11533 2.2 Total 35 260

NDVI in Established Camp Zones Most camp zones are well established and have been used for decades and are occupied every year by multiple camps, whereas others were occupied only once or twice during our study. We found that those areas that were occupied year after year have, on average, higher NDVI values than the areas that were never occupied and those areas that were occupied only once or twice during our study (see ﬁg. A2). One-way Analysis of Variance (ANOVA) comparing the number of years that an area was occupied shows that there are statistically signiﬁcant differences between the cells that were occupied for longer periods of time (P !.0001). Thus, the camp zones that are most frequently used represent habitats that have over the years consistently higher NDVI values, that is, more grazing resources.

Figure A2. Average NDVI value and years of occupation of camp zones.

3 Supplement A from Moritz et al., Mobile Pastoralists in the Logone Floodplain

Spatial Distributions of Mobile Pastoralists and NDVI We examined whether there was evidence for an ideal free distribution in the Logone floodplain by testing whether there was a positive correlation between the total NDVI and the total number of animals per camp zone for the 5 different years. We also examined whether there were any patterns in overuse and under use, that is, whether the percentage of NDVI per camp zone buffer (as a percentage of the NDVI of all the camp zone buffers) was equal to the percentage of animals per camp zone buffer (as a percentage of the animals in all the camp zone buffers) in each of the years. If the percentage of animals was higher than the percentage of NDVI, it is considered a case of overuse (more animals than resources). If the percentage of animals was lower that the percentage of NDVI, it is considered under use (fewer animals than resources). We found that the number of camp zones that were overused stayed more or less the same across the years, although it was not always the same camp zones that were overused (see figs. A4–A8 and their legend, fig. A3).

Figure A3. Legend for ﬁgs. A4–A8.

4 Supplement A from Moritz et al., Mobile Pastoralists in the Logone Floodplain

Figure A4. Distribution of mobile pastoralists in the study area in February 2008. For legend, see ﬁg. A3.

5 Supplement A from Moritz et al., Mobile Pastoralists in the Logone Floodplain

Figure A5. Distribution of mobile pastoralists in the study area in February 2009. For legend, see ﬁg. A3.

6 Supplement A from Moritz et al., Mobile Pastoralists in the Logone Floodplain

Figure A6. Distribution of mobile pastoralists in the study area in February 2010. For legend, see ﬁg. A3.

7 Supplement A from Moritz et al., Mobile Pastoralists in the Logone Floodplain

Figure A7. Distribution of mobile pastoralists in the study area in February 2011. For legend, see ﬁg. A3.

8 Supplement A from Moritz et al., Mobile Pastoralists in the Logone Floodplain

Figure A8. Distribution of mobile pastoralists in the study area in February 2012. For legend, see ﬁg. A3.

9 Supplement A from Moritz et al., Mobile Pastoralists in the Logone Floodplain

References Cited Only in Supplement A Allee, W. C., A. E. Emerson, O. Park, T. Park, and K. P. Schmidt. 1949. Principles of animal ecology. Philadelphia: Saunders. Arditi, R., and B. Dacorogna. 1988. Optimal foraging on arbitrary food distributions and the definition of habitat patches. American Naturalist 131:837–846. Cantrell, R. S., C. Cosner, and Y. Lou. 2008. Approximating the ideal free distribution via reaction-diffusion-advection equations. Journal of Differential Equations 245:3687–3703. Cressman, R., and V. Krivan. 2006. Migration dynamics for the ideal free distribution. American Naturalist 168:384–397. Delclaux, F., C. Seignobos, G. Lieńou, and P. Genthon. 2010. Water and people in the Yae´re´ floodplain (North Cameroon). In Floodplains: physical geography, ecology and societal interactions.M.A.A´ lvarez, ed. Pp. 1–27. Hauppauge, NY: Nova. Jung, H. C., D. Alsdorf, M. Moritz, H. Lee, and S. Vassolo. 2011. Analysis of the relationship between flooding area and water height in the Logone floodplain. Physics and Chemistry of the Earth 36:232–240. Kelly, R. L. 1995. The foraging spectrum: diversity in hunter-gatherer lifeways. Washington, DC: Smithsonian Institution Press. Nicholson, M. J., and M. H. Butterworth. 1986. A guide to condition scoring of Zebu cattle. Addis Ababa, Ethiopia: International Livestock Centre for Africa. Winterhalder, B., and D. J. Kennett. 2006. Behavioral ecology and the transition from hunting and gathering to agriculture. In Behavioral ecology and the transition to agriculture. D. J. Kennett and B. Winterhalder, eds. Pp. 1–21. Berkeley: University of California Press.